CDN Performance Benchmarking: Methodology and Tools

September 13, 2026 · Alex Petrov · 4 min read

Why Benchmark CDNs

CDN performance varies by: geographic region (a CDN fast in the US may be slow in Southeast Asia), content type (static vs. dynamic), time of day (congestion patterns), and configuration (your specific cache rules and origin setup). Generic "fastest CDN" claims are meaningless without context. Benchmark with your own content, from your users' locations. For foundational context, read CDN Monitoring and Observability: Metrics That Matter — FlyFish NWA.

Benchmarking Methodology

Step 1: Define What You're Measuring

TTFB — Time to first byte. Measures CDN's processing + network latency.
Full page load — Complete page render time. Includes all assets.
LCP — Largest Contentful Paint. Google's key Core Web Vitals metric.
Cache hit latency vs. miss latency — How fast cache hits are vs. cache misses.
Purge propagation time — How quickly cache invalidation takes effect globally.

Step 2: Choose Testing Locations

Test from locations where your users are. A CDN optimized for North America won't help if 40% of your users are in Asia.

Performance Comparison: Edge vs Origin Processing
Metric	Edge Processing	Origin Processing	Improvement
Latency (p50)	8–15 ms	120–300 ms	15–20x faster
Cache Hit Ratio	85–95%	N/A	Reduces origin load
TLS Handshake	1–5 ms (nearby PoP)	50–150 ms	10–30x faster
Bandwidth Cost	$0.01–0.04/GB	$0.05–0.12/GB	3–5x cheaper
Availability	99.99% (multi-PoP)	99.9% (single DC)	10x fewer outages

Figure 1: CDN Performance Dashboard — Core Web Vitals with CDN optimization

If your audience is:
60% US → Test from 5 US cities (East + West + Central)
20% Europe → Test from London, Frankfurt, Paris
15% Asia → Test from Tokyo, Singapore, Mumbai
5% Other → Test from São Paulo, Sydney

Step 3: Test Methodology

Run tests for at least 24 hours (catches time-of-day variations)
Test the same content across all CDN providers
Include cache-hit AND cache-miss scenarios
Test during normal traffic AND peak hours
Run at least 100 requests per location per hour

Testing Tools

Tool	Type	Best For
Catchpoint	Synthetic	Enterprise-grade, 800+ locations
WebPageTest	Synthetic	Free, detailed waterfall analysis
Datadog Synthetics	Synthetic	Integration with APM stack
curl + cron	DIY	Simple, free, customizable

Simple DIY Benchmark Script

#!/bin/bash
# Test TTFB from current location
for cdn in cloudflare cloudfront fastly; do
  echo "=== $cdn ==="
  for i in $(seq 1 100); do
    curl -o /dev/null -s -w "%{time_starttransfer}
"       "https://$cdn-test.example.com/benchmark.jpg"
  done | awk '{sum+=$1} END {print "Avg TTFB:", sum/NR, "s"}'
done

Interpreting Results

Look at percentiles, not averages. P50 shows typical experience; P95 shows tail latency; P99 shows worst-case. A CDN with 20ms P50 but 500ms P99 has an intermittent problem that averages hide.

Figure 2: Request Timing — Latency breakdown across the CDN stack

Common Pitfalls

Testing only from one location (not representative)
Testing only cache hits (miss latency matters too)
Short test duration (misses time-of-day patterns)
Not matching real traffic patterns (test with representative content)